# Seedance 2.5 Set to Expand Multimodal Input Support to 50 Media Files > Source: > Published: 2026-07-30 16:58:57+00:00 *The next version of Seedance is expected to let creators feed dozens of reference files into a single generation task rather than working one input at a time.* Coming soon to the [Seedance AI video platform](https://www.seedance2ai.io/), Seedance 2.5 is set to support up to 50 media inputs within a single creative task. The reference stack can include up to 44 images, three video clips and three audio files, bringing visual, motion and sound references into one multimodal workflow. The tool is not yet fully available, and the team behind it describes the update as still in progress ahead of a wider rollout. ## A larger reference stack The update builds on the multimodal foundation introduced in Seedance 2.0, with a larger reference stack, longer video generation, higher-resolution output and what the team describes as improved creative control. Where earlier versions generally worked from a narrower set of reference materials, Seedance 2.5 is designed to let creators combine multiple input types in one task, including character images, product references, scene examples, motion clips, camera references and music or other audio materials. Combining that range of material in a single workflow is intended to give creators more control over characters, scenes, movement, sound, visual style, camera direction and overall continuity throughout a generated clip. It could also reduce the need to manage different media types across separate tools or production stages, a common friction point in AI video workflows that typically require moving between distinct generation and editing tools for visuals, motion and audio. For teams producing a series of clips around the same character or product, being able to draw on that many reference points at once is meant to cut down on the back and forth of re-uploading and re-aligning materials for every new generation. ## Longer clips, higher resolution Seedance 2.5 is also expected to generate videos of up to 30 seconds in a single pass, along with native 4K output. The team points to improved scene and character consistency and camera movement as part of the update, along with 3D spatial pre-visualization and camera-blocking tools aimed at planning shots and camera paths before full generation. Planning a shot in three dimensions before committing to a full render is intended to give creators a way to check framing, movement and pacing early, rather than discovering an issue only after a clip has already been generated. Taken together, the changes are aimed at workflows where creators need a tool that can hold onto a consistent character or scene across a longer clip rather than a single short cut. Potential applications include product videos, advertising, social media content and film pre-visualization, alongside general creator workflows where assembling references from multiple sources is typically slow. The team was clear that comparisons to other video generation models should be read with some caution, and Seedance 2.5 is not presented as universally ahead of competing tools in the space. Users remain responsible for securing rights to any commercial, branded or licensed materials they choose to use as inputs. More details on the platform, including how the expanded input support works in practice, are available at their site, seedance2ai.io.